|
From: Ian P. <ian...@in...> - 2004-04-08 16:16:28
|
On 08 Apr 2004, at 18:05, Andreas Raab wrote: > Incidentally, talking about message forwarding. We might want to keep > in > mind the stuff that Stephen Pair did for the Chango VM - is he on this > list? > If not, we should invite him. He had some fairly interesting stuff > going > with delegation and that might be of some interest here. Which reminds me, for V4 the following might be on my wishlist too: - separate the notion of "self" and "receiver". For most purposes they remain the same thing. Within #doesNotUnderstand: (or in some kind of explicit primitive #perform: type thing that allows divergence) they can be made different objects. Receiver inst vars remain with the "receiver" definition, while sends to self remain with the "self" definition. This allows (for example) "bullet-proof proxies" where #dNU: sets receiver to the object that fields the send and retains the proxy as "self". Any subsequent send to self goes straight back to the proxy, rather than the receiver, without affecting how state is accessed within the receiver; i.e., it's completely transparent to the (non-meta ;) programmer. I think VisualWorks does (or did or was going to do a long time ago) something similar, precisely for the proxy interest. Cheers, Ian |